Runway: a voice becomes an object, and then has an id
Creation is asynchronous. A voice moves through processing into a ready state, arrives with a preview to listen to, and is afterwards addressed by id. Auditioning happens once, before any shot exists. As of 2026-09-22.
| What the documentation settles | What it leaves to a take |
|---|---|
| A voice is created asynchronously and reaches a ready state with a preview | Whether the preview predicts how the voice performs a whole scene |
| The voice is then referred to by its own id | Whether an id can be re-pointed at a different sample |
| A custom voice comes from a 10-second to 5-minute sample, or a description of 20 characters | Whether a described voice is reproducible from the same sentence |
Inclusion rule. Only statements found on the vendor page named in the caption, and only ones bearing on this one field. Nothing here is filled in from generated output. Order. In the order the statements arrive on the vendor's page.
1A preview is the part nothing else here offers
Across this register, voices are bound, supplied, catalogued or unmentioned, and only this one is auditioned as a documented step. A voice that can be heard and approved before any generation turns casting into a decision rather than a discovery.
The asynchronous lifecycle is what makes that possible. Something has to exist between the request and the shot for there to be anything to listen to, and a ready state with a preview is exactly that something.
2An object with an id is the durable form, whoever holds it
This is a voice object rather than a character object, so the pairing between a person in the story and a voice still lives in a production's own records. The difference from a bound-to-the-character arrangement is where the bookkeeping sits, not whether it exists.
In exchange the voice can be reused anywhere, shared between characters deliberately, or retired without touching the character. That flexibility is the reason a voice object is usually the better trade for a production with more than one series.
3The others where a voice exists before any shot does
Three more entries let a voice outlast a call. Two attach it to a character; one enrols a clone and reuses it by id, which is the closest arrangement to this one.
- HeyGen — on a stored clone.
- Kling AI — on the element.
- LTX Studio — on the character element.
- Per-character bindingA voice is created asynchronously, reaches a ready state with a preview, and is then referred to by its own ida stored voice object
- Voice sourceA custom voice can be built from an audio sample between 10 seconds and 5 minutes long and at most 10 MBa published sample window
- Voice sourceA voice can instead be asked for in words, with the description required to run at least 20 charactersa route that needs no recording
4Sources
Read from the custom voices reference at docs.dev.runwayml.com on 2026-09-22. The same column across every entry is on per-character binding; everything this vendor publishes about speech is on Runway. What counts as documented is on how read.